为了提高人脸识别系统识别率和算法的运行效率,提出一种依据阈值变化的自适应定位人眼瞳孔的方法.该方法根据最小极区的思想,利用阈值从低到高的动态变化,借助自然约束条件、相似度分析和灰度重心法来精确定位人眼瞳孔的位置.利用该方法在CAS_PEAL、CVL以及自建的人脸库上进行了大量实验.实验结果表明,该人眼瞳孔定位方法可以有效定位人眼瞳孔,并且对于光照、表情都具有较好的鲁棒性.
A practical face recognition system makes high demands on accuracy and real-time.The eye-pupil locating is a key factor which effects recognition rate and efficiency.In this paper,an eye-pupil loacating method based on adaptive threshold changes is proposed.The eye pupils are located by the threshold changing,similarity analysis,natural constraints condition and gray centre of gravity according to the idea of MER(minimum extremal region).A lot of experiments are done on CAS_PEAL,CVL and self-construct face database.The experimental results show that the method is effective and has a high locating rate.And the method is robust for variant illumination and face expression.